rotnet

RotNet

input

(from https://github.com/d4nst/RotNet/tree/master/data/test_examples)

Ailia input shape: (1, 224, 224, 3)

output

• Original: original image (after cropped)
• Rotated: input image (randomly rotated)
• Corrected: output image (model output is predicted angle, therefore we rotated the "rotated image" to visualize our output)

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 rotnet.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 rotnet.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

By adding the --video option, you can input the video.
If you pass 0 as an argument to VIDEO_PATH, you can use the webcam input instead of the video file.

$ python3 rotnet.py --video VIDEO_PATH

Currectly, two pretrained-models are avilable:

• mnist(for mnist dataset)
• gsv2(for google street view dataset) You can select one of them by adding --model (default: gsv2).

Reference

CNNs for predicting the rotation angle of an image to correct its orientation

Framework

Keras

Model Format

ONNX opset = 10

Netron

rotnet_gsv_2.onnx.prototxt